It was the version number heard ’round the world. The momentous first release of ChatGPT on Nov. 30, 2023 was based on GPT 3.5, the nonzero number firmly establishing the skyward trajectory of artificial intelligence. That's what makes it the Emerge 2023 Project of the Year.
ChatGPT changed everything
There was no doubt that ChatGPT was the most consequential project of the year. Even including the other sectors that Decrypt currently tracks under Emerge—biohacking and longevity, quantum computing, space 2.0 and beyond—the software release was a pivotal one.
As 2023 draws to a close, entire industries already appear to be built on top of ChatGPT, sometimes at their peril. The tool—largely through its application programming interface (API)—has been woven into the fabric of tens and thousands of companies. Every single tech product has an AI feature in the works, if not already released.
It's difficult not to describe it in hyperbolic terms.
In barely over a year, ChatGPT rocketed from obscure, largely academic roots to the embodiment of a revolution that has rattled global governments and captured the attention of every major enterprise and industry on Earth. There have been proclamations from the United Nations, international treaties, and congressional hearings. There have been national bans and prohibitions on its use in government, the military, and the private sector.
Perhaps most tellingly, even people behind the technology are calling for its regulation. ChatGPT and AI are embedding themselves in systems everywhere so fast that it's scaring some of the smartest minds on the planet.
In May, hundreds of tech and business leaders signed an open letter declaring that there is “risk of extinction from AI,” and that controlling it should be as important as nuclear war and global pandemics.
ChatGPT fingerprints are already everywhere, and as revolutions go, we're just getting started.
What's in a name?
To be sure, it's not the sexiest branding, As project—and product—names go, ChatGPT is borderline generic.
“GPT” is a widely used acronym in the field of machine learning, classifying the tool as one based on a generative pre-trained transformer (GPT). And that acronym is only mid-way through a recursive chain of other acronyms that the world has had to quickly learn, including LLM (large language model) and NLP (natural language processing).
It wasn't even new.
The first commercially viable version of ChatGPT was based on GPT 3, the third generation of the technology that OpenAI introduced in 2018. And that technology used generative pre-training, an implementation of AI model building that dates back to 2012.
The "transformer" piece didn't arrive until 2017 thanks to employees at Google, who proposed it in a research paper. The following year, OpenAI published its own paper, "Improving Language Understanding by Generative Pre-Training." For the then non-profit OpenAI, the stage had been set.
Nothing but a number
It was ChatGPT 3.5 that software developers could access with OpenAI's first API. ChatGPT 3.5 was also the basis of OpenAI's core mobile apps for iOS and Android.
When the next major public upgrade to ChatGPT was announced, it didn't have a cool codename like MacOS Leopard or Windows XP. The world was introduced to the tool as ChatGPT 3.5, and the subsequent sequential AI model was dubbed ChatGPT 4.
When it was unveiled in March, the world took notice, the update revealing that a global game changer was also a fast-moving juggernaut. OpenAI made clear it was pushing forward hard. So hard, it turns out, that it scared people.
It was GPT-4 that gave people the first sense of velocity, that set the pace of change, that turned a major point in history into a tidal wave of change. With a prominent petition signed by prominent names urging a pause in development, Sam Altman ended up having to reassure the world that OpenAI was going to focus on making ChatGPT 4 better (and safer), rather than barreling ahead to ChatGPT 5.
But last month, in what was a surprise to no one, Altman acknowledged that ChatGPT 5 was indeed in the works. And we can probably guess what subsequent versions will be called.
Given the monumental impact of the already dated ChatGPT 3.5, however, it's almost impossible to imagine what AI will look like in another year.
Edited by Guillermo Jimenez.